KNIGHTBYRD VERIS · CLAIM EVIDENCE
Your marketing makes claims.
What backs each one?
VERIS reads a page, records the claims it finds, verbatim, and checks whether the evidence published beside each one can actually carry it. Dated at the moment it was read.
Start with your own page. Free, no account, under a minute.
One public page. Nothing is stored about you, and no account is needed.
Already know what you need? See the six products and prices
WE WENT FIRST
We ran it on ourselves before we ran it on anyone else. It found two of our own claims that would face the same question — and we published them.
“Most homeowners overpay on property taxes.” — a factual assertion with nothing cited beside it.
“…built by experienced operators who have been working inside government, financial services, food retail…” — sector breadth with no disclosed boundary.
Found on knightbyrd.com, 19 Aug 2026. We would rather tell you that than scan everyone except ourselves.
EVIDENCE DECAY
Your claim did not change. The study cited beside it did.
This is the failure no single reading of a page can find, and the one page monitors are built to miss. They watch your page. When nothing on your page moves, they stay quiet — and nothing on your page moves when the source behind a claim is rewritten, redirected, or quietly taken down.
VERIS re-reads the sources your claims cite, not only the claims, and reports five states: live, moved, changed, gone, unreachable. A source we could not reach is reported as unreachable — never as missing. Seeing it at all requires two dated readings, which is why it arrives with a subscription and not with a scan.
Continuous re-reading of cited sources is not described in published vendor materials for any category compared below, as of 19 Aug 2026. That is a statement about what those vendors document — not a claim that it cannot be done elsewhere.
What it found in one run
16 companies read on 19 Aug 2026. Four more were attempted and unreadable — excluded from the figures below rather than counted as clean.
Two came back clear, and that is the part worth reading. A tool that flags nothing is broken. A tool that flags everything is worthless. This one separates them.
What the category does, and what it does not
We compare by product category, never by company. A dash means a capability is not described in published vendor materials as of 19 Aug 2026 — not that it is impossible. That distinction is the whole product, so we apply it here too.
| Capability | Change-detection platforms $16k–$100k/yr · days to weeks | Enterprise compliance suites six figures · 8–16 week setup | SMB review tools from ~$160/mo · same day | Ad-creative archives varies · same day | VERIS |
|---|---|---|---|---|---|
Reduces a page to its individual claims A substantiation request is never about a page. It is about one sentence, and someone has to find it first. | Each claim it finds, recorded verbatim and classified across ten kinds. | ||||
Says whether the cited source can carry the claim Evidence being present is not evidence being sufficient. | A table of what each source class can and cannot carry, checked against the exact assertion. | ||||
Detects breadth claimed vs breadth tested The claim was accurate for what was tested and unsupported for what was promised. That is the pattern regulators have acted on. | Unstated tested scope stays UNKNOWN — never assumed to be zero. | ||||
Dated provenance on each line Substantiation is judged as of the date the claim was made. | Each artifact carries the date it was fetched. The quality gate refuses to render one without. | ||||
Watches what happens to the sources a claim cites The expensive failure is the page NOT moving while the study cited beside the claim quietly stopped resolving. | Five evidence states, re-read on every sweep. Unreachable is reported as unreachable, never as missing. | ||||
Tests both disclosure requirements separately One label does not satisfy two rules. | Catches the trap in both directions, and reports UNKNOWN rather than guessing. | ||||
Points the same inventory at a competitor Find out what a rival published to support the claim that just beat you — as a record, not an accusation. | The identical inventory, pointed the other way. Accusatory output is rejected by our own gate. | ||||
Time to a first result The difference between finding out now and finding out after someone asks. | Under a minute. Paste a URL, no account. | ||||
Publishes its own results against itself Anyone can grade someone else. Ask what the tool says about the company selling it. | We scanned knightbyrd.com and published the two findings it returned against us. |
Evidence present is not evidence sufficient
A review-platform ranking substantiates that people liked your product. It cannot substantiate a measured performance claim — a review site does not measure performance. The citation looks like proof right up until someone asks what it measured.
Breadth claimed is not breadth tested
A claim can be accurate for the set you measured and unsupported for the set you promised. It reads fine in the copy, and it is the pattern that has actually drawn enforcement.
The page stops changing. The product does not
Change monitoring fires when a page moves. The expensive failure is the opposite: nothing on the page has moved while the product shipped four versions past the evidence behind it.
Who it is for
Your marketing makes claims. What backs each one?
The claims found on the pages you submit, inventoried and dated, with the evidence cited beside each one and whether that source carries the assertion made.
Buy the LedgerYou lost the deal to a claim that sounded stronger than yours.
A dated inventory of a named competitor's public claims and the evidence published alongside each one. It records what was published; it draws no conclusion.
Buy the BlazonSubstantiation requests do not come with notice.
Your substantiation file assembled before anyone asks: each published claim we find, its cited source, the date that source was produced, and where the file is thin.
Buy the DossierYour claim did not change. The source behind it stopped resolving.
Up to ten of your own pages, re-read daily, with a private dated record: what changed, how long each claim has stood unchanged, what happened to the sources you cite, and the substantiation question each change raises. You hear from us only when something material moves.
Buy the SentryTheir claim did not change. The study behind it did.
Daily reading of one named competitor's public claims and of the sources those claims cite. One competitor per subscription. It reaches no conclusion about them — what the record means is your call.
Buy the Blazon WatchA machine cannot tell your careful claim from an invented one.
A dated record of the claims found on a page and the evidence published beside them, at a permanent public address — a page a person can read and a machine-readable feed an agent can fetch, generated from one reading so the two cannot disagree.
Buy the AttestationEach one starts with the free scan above. Setup is a URL — the category standard for this capability is an 8–16 week implementation. Terms, refunds and scope: how we work.
What VERIS does not do
- It does not determine whether a claim is adequate.
- It does not give legal advice or reach a legal conclusion.
- It does not write a claim you did not publish. Nothing is paraphrased — each line appears as published.
- It does not treat an unknown as a zero. If a tested scope was never stated, it says so.
- It does not accuse anyone. Accusatory output is rejected by its own quality gate.
Common questions
What counts as a claim?
A sentence your marketing publishes that asserts something checkable — an accuracy figure, a comparison, a coverage promise, a guarantee, or a capability. VERIS records them verbatim; it never writes a claim you did not publish.
Is evidence being present the same as evidence being sufficient?
No, and the gap between them is where claims get challenged. A review-platform ranking substantiates that people liked your product. It cannot substantiate a measured performance claim, because a review site does not measure performance.
What is a scope mismatch?
Breadth claimed versus breadth tested. A claim can be accurate for what was measured and unsupported for what was promised — a detector trained on academic writing but marketed as working on all content types — tested narrowly, sold broadly. That gap is what this reads for.
Does VERIS tell me whether I am compliant?
No. VERIS records what was claimed and what evidence was published alongside it. It does not determine whether a claim is adequate, and it is not legal advice or a legal conclusion.
Does one disclosure label cover both an ad and AI involvement?
They are separate requirements. A commercial-relationship label does not satisfy an AI-involvement requirement, and an AI label does not satisfy the commercial one. Marking one and assuming both are covered is the common failure.
How fast is the first result?
Under a minute. Paste a URL and the free scan returns a dated count, with no account and nothing installed.
How many claims are on your homepage right now?
If the answer is a guess, that is the problem. The scan is free and takes under a minute.
VERIS is an evidence inventory. It records what was claimed and what evidence was published alongside it. It does not determine whether a claim is adequate, and it is not legal advice or a legal conclusion. Category comparisons describe published vendor materials as of 19 Aug 2026 and name no company. Counts come from the visible text of the page submitted.